CN112348796A - Cerebral hemorrhage segmentation method and system based on combination of multiple models - Google Patents
Cerebral hemorrhage segmentation method and system based on combination of multiple models Download PDFInfo
- Publication number
- CN112348796A CN112348796A CN202011235586.1A CN202011235586A CN112348796A CN 112348796 A CN112348796 A CN 112348796A CN 202011235586 A CN202011235586 A CN 202011235586A CN 112348796 A CN112348796 A CN 112348796A
- Authority
- CN
- China
- Prior art keywords
- output result
- image data
- cerebral hemorrhage
- convolution
- size
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 206010008111 Cerebral haemorrhage Diseases 0.000 title claims abstract description 32
- 230000011218 segmentation Effects 0.000 title claims abstract description 30
- 210000004556 brain Anatomy 0.000 claims abstract description 43
- 208000032843 Hemorrhage Diseases 0.000 claims abstract description 18
- 230000000740 bleeding effect Effects 0.000 claims abstract description 18
- 210000003625 skull Anatomy 0.000 claims abstract description 18
- 238000001514 detection method Methods 0.000 claims abstract description 17
- 230000000877 morphologic effect Effects 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims description 31
- 230000006870 function Effects 0.000 claims description 28
- 238000012360 testing method Methods 0.000 claims description 26
- 230000004913 activation Effects 0.000 claims description 21
- 238000003062 neural network model Methods 0.000 claims description 13
- 210000001175 cerebrospinal fluid Anatomy 0.000 claims description 6
- 230000003628 erosive effect Effects 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 abstract description 4
- 208000006011 Stroke Diseases 0.000 description 6
- 208000016988 Hemorrhagic Stroke Diseases 0.000 description 5
- 230000006872 improvement Effects 0.000 description 4
- 208000020658 intracerebral hemorrhage Diseases 0.000 description 4
- 238000011176 pooling Methods 0.000 description 4
- 208000008574 Intracranial Hemorrhages Diseases 0.000 description 3
- 230000034994 death Effects 0.000 description 3
- 231100000517 death Toxicity 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000010339 dilation Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 208000032382 Ischaemic stroke Diseases 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000007797 corrosion Effects 0.000 description 2
- 238000005260 corrosion Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 210000004958 brain cell Anatomy 0.000 description 1
- 210000005013 brain tissue Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 210000001519 tissue Anatomy 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/155—Segmentation; Edge detection involving morphological operators
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30016—Brain
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention provides a cerebral hemorrhage segmentation method and a system based on multi-model combination, wherein the method comprises the following steps: acquiring brain image data to be detected; removing a skull part in the brain image data to be detected by adopting a morphological method to obtain preprocessed data; inputting the preprocessed data into a target detection model, and outputting a segmentation result of the brain image data, wherein the segmentation result comprises a brain bleeding area. Therefore, the bleeding area of the brain image can be accurately segmented, the image processing efficiency is higher, a doctor can conveniently find out the bleeding area of the brain rapidly, and the success rate of the operation is improved.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a cerebral hemorrhage segmentation method and system based on multi-model combination.
Background
Stroke is the second most common cause of death and disability worldwide, with nearly 80 million people experiencing new or recurrent strokes each year, survivors beyond 2/3 will have some type of disability, and the "gold phase" of stroke treatment is 3 to 6 hours after onset. Strokes are mainly classified into two types, ischemic strokes and hemorrhagic strokes. Hemorrhagic stroke accounts for approximately 13% of stroke cases and 40% of all stroke deaths. Hemorrhagic stroke is bleeding of the surrounding brain tissue (cerebral hemorrhage) caused by rupture of cerebral blood vessels, resulting in death of brain cells and the affected parts of the brain will stop working. Hemorrhagic stroke has a higher mortality rate and greater risk than ischemic stroke.
At present, cerebral hemorrhage is treated, and the cerebral hemorrhage is generally segmented on a head X-ray CT image. However, this segmentation is performed manually through the intervention of a professional, and is complicated in operation and requires a high experience of a doctor.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a cerebral hemorrhage segmentation method and system based on multi-model combination.
In a first aspect, the present invention provides a method for segmenting cerebral hemorrhage based on multi-model combination, including:
step 1: acquiring brain image data to be detected;
step 2: removing a skull part in the brain image data to be detected by adopting a morphological method to obtain preprocessed data;
and step 3: inputting the preprocessed data into a target detection model, and outputting a segmentation result of the brain image data, wherein the segmentation result comprises a brain bleeding area.
Optionally, before performing step 3, the method further includes:
constructing an image dataset comprising a cerebral hemorrhage region;
dividing the image data set into a training set and a test set;
classifying the images in the training set and the test set respectively based on a fuzzy C-means clustering method to obtain intermediate image data marked with image region classification results;
removing the skull part in the intermediate image data by adopting a morphological method to obtain target data;
training the constructed initial detection model through the target data corresponding to the training set to obtain a trained model; the initial detection model is an improved U-Net neural network model established by combining hole convolution;
and testing the trained model through the target data corresponding to the test set, and if the test is passed, obtaining the target detection model.
Optionally, the constructing an image data set containing a cerebral hemorrhage region includes:
and screening the images containing the cerebral hemorrhage area, screening the images with the image size, the image definition and the image integrity meeting the requirements, and adding the images into the image data set.
Optionally, classifying the images in the training set and the test set respectively based on a fuzzy C-means clustering method to obtain intermediate image data labeled with image region classification results, including:
based on a fuzzy C-means clustering method, clustering is respectively carried out on the gray value of each pixel of the images in the training set and the testing set, and intermediate image data marked with four types of region classification results, namely a grey brain matter region, a white brain matter region, a cerebrospinal fluid region and a bleeding region, are obtained.
Optionally, removing the skull portion in the intermediate image data by using a morphological method to obtain target data, including:
and adjusting the expansion parameter and the erosion parameter of the intermediate image data until the skull part in each image is removed.
Optionally, the building of the improved U-Net neural network model in combination with the hole convolution includes a seven-layer convolution structure, wherein:
the first layer is used for: performing two convolutions and a cavity convolution on input data, wherein the size of a convolution kernel is 3 multiplied by 3, an activation function is relu, stride is 2, and a filling mode is nonpadding to obtain a first output result;
the second layer is used for: performing two convolutions and a cavity convolution on the first output result, wherein the convolution kernel size is 3 multiplied by 3, the activation function is relu, stride is 2, and the filling mode is nonpadding, so as to obtain a second output result;
the third layer is used for: performing two convolutions and a cavity convolution on the second output result, wherein the convolution kernel size is 3 multiplied by 3, the activation function is relu, stride is 2, and the filling mode is nonpadding, so as to obtain a third output result;
the fourth layer is used for: performing two convolutions on the third output result, wherein the size of a convolution kernel is 3 multiplied by 3, an activation function is relu, and a filling mode is nonpadding to obtain a fourth output result;
the fifth layer is used for: performing convolution and clipping operations on the copied first output result, the copied second output result and the copied third output result, then performing splicing and upsampling operations, wherein the size of a convolution kernel is 3 multiplied by 3, and the size of upsampling is 2 multiplied by 2, so as to obtain a fifth output result;
the sixth layer is for: performing two convolution and one upsampling on the copied fourth output result and the fifth output result, wherein the size of a convolution kernel is 3 multiplied by 3, the size of an activation function is relu, the size of upsampling is 2 multiplied by 2, and the filling mode is nopadding to obtain a sixth output result;
the seventh layer is for: performing two convolutions on the copied fifth output result and the copied sixth output result, wherein the size of a convolution kernel is 3 multiplied by 3, an activation function is relu, and a filling mode is nonpadding to obtain a seventh output result;
and performing 1 × 1 convolution on the seventh output result, wherein the activation function is relu, and the segmentation result of the brain image data is obtained.
Optionally, when the initial detection model constructed by training the target data corresponding to the training set is used, the influence on the training result is observed by using three loss functions, namely, binarization cross entropy, dice loss and focus loss.
In a second aspect, the present invention provides a brain hemorrhage segmentation system based on multi-model combination, which includes a memory, and a processor, where the memory stores a computer program, and when the processor calls the computer program stored in the memory, the brain hemorrhage segmentation system based on multi-model combination according to any one of the first aspect is performed.
Compared with the prior art, the invention has the following beneficial effects:
the invention provides a cerebral hemorrhage segmentation method and system based on multi-model combination, which are characterized in that brain image data to be detected are obtained; removing a skull part in the brain image data to be detected by adopting a morphological method to obtain preprocessed data; inputting the preprocessed data into a target detection model, and outputting a segmentation result of the brain image data, wherein the segmentation result comprises a brain bleeding area. Therefore, the bleeding area of the brain image can be accurately segmented, the image processing efficiency is higher, a doctor can conveniently find out the bleeding area of the brain rapidly, and the success rate of the operation is improved.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
fig. 1 is a flowchart of a method for segmenting cerebral hemorrhage based on multi-model combination according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an improvement of a U-Net neural network model provided by an embodiment of the present invention.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
The invention provides a cerebral hemorrhage segmentation method based on multi-model combination, which comprises the following steps: step 1: acquiring brain image data to be detected; step 2: removing a skull part in the brain image data to be detected by adopting a morphological method to obtain preprocessed data; and step 3: inputting the preprocessed data into a target detection model, and outputting a segmentation result of the brain image data, wherein the segmentation result comprises a brain bleeding area.
The method provided by the invention can accurately segment the bleeding area of the brain image, has higher image processing efficiency, is convenient for doctors to quickly find out the bleeding area of the brain, and improves the success rate of the operation.
Illustratively, before step 3 is executed, an object detection model needs to be constructed, specifically, the following steps are included: constructing an image dataset comprising a cerebral hemorrhage region; dividing an image data set into a training set and a test set; classifying the images in the training set and the test set respectively based on a fuzzy C-means clustering method to obtain intermediate image data marked with image region classification results; removing the skull part in the intermediate image data by adopting a morphological method to obtain target data; training the constructed initial detection model through target data corresponding to the training set to obtain a trained model; the initial detection model is an improved U-Net neural network model established by combining hole convolution; and testing the trained model through the target data corresponding to the test set, and if the test is passed, obtaining a target detection model.
Fig. 1 is a flowchart of a method for segmenting cerebral hemorrhage based on multi-model combination according to an embodiment of the present invention, as shown in fig. 1, the method in this embodiment may include:
step (a): collection of data required for the experiment.
Illustratively, in step (a), a data set of cerebral hemorrhage images is obtained first, the size, definition and integrity of each image data is checked, image data suitable for experiment is selected, the size of the final data set is confirmed, and the final data set is divided into a training set and a testing set.
Step (b): the collected data were clustered into 4 classes of grey brain matter, white brain matter, cerebrospinal fluid and bleeding area by FCM method.
Illustratively, in step (b), the brain hemorrhage image is clustered with the FCM algorithm for the gray value of each pixel. For example, the parameters c are 4, and the images are clustered into 4 classes, i.e., a gray brain matter region, a white brain matter region, a cerebrospinal fluid region, and a bleeding region.
In this embodiment, the FCM method is used to pre-segment the image. The craniocerebral CT image has uncertainty and ambiguity, clear boundaries among different tissues are difficult to find, and the FCM clustering algorithm is suitable for the characteristics of uncertainty and ambiguity in gray level images, so that the boundaries are clearer and the subsequent skull removal and segmentation operations are facilitated.
Step (c): the skull is removed by morphological image method dilation and erosion.
Illustratively, in step (c), the skull portion is removed from the clustering result in step (b) by morphological dilation and erosion, and then parameters of dilation and erosion are adjusted until the skull of each image is removed.
Step (d): and establishing an improved U-Net neural network model by combining the hole convolution.
Fig. 2 is a schematic diagram of an improvement of a U-Net neural network model provided in an embodiment of the present invention, and referring to fig. 2, the improved U-Net neural network model built by combining a hole convolution includes a seven-layer convolution structure, in which:
the first layer is used for: and performing two convolutions and a cavity convolution on the input data, wherein the size of a convolution kernel is 3 multiplied by 3, an activation function is relu, stride is 2, and a filling mode is nonpadding, so as to obtain a first output result.
The second layer is used for: and performing two convolutions and a hollow convolution on the first output result, wherein the convolution kernel size is 3 multiplied by 3, the activation function is relu, stride is 2, and the filling mode is nonpadding, so as to obtain a second output result.
The third layer is used for: and performing two convolutions and a hollow convolution on the second output result, wherein the convolution kernel size is 3 multiplied by 3, the activation function is relu, stride is 2, and the filling mode is nonpadding, so as to obtain a third output result.
The fourth layer is used for: and performing two convolutions on the third output result, wherein the size of a convolution kernel is 3 multiplied by 3, the activation function is relu, and the filling mode is nopadding, so as to obtain a fourth output result.
The fifth layer is used for: and performing convolution and clipping operations on the copied first output result, the copied second output result and the copied third output result, then performing splicing and upsampling operations, wherein the size of a convolution kernel is 3 multiplied by 3, and the size of upsampling is 2 multiplied by 2, so that a fifth output result is obtained.
The sixth layer is for: and performing two convolution and one upsampling on the copied fourth output result and the fifth output result, wherein the size of a convolution kernel is 3 multiplied by 3, the size of an activation function is relu, the size of upsampling is 2 multiplied by 2, and the filling mode is nopadding, so that a sixth output result is obtained.
The seventh layer is for: and performing two convolutions on the copied fifth output result and the sixth output result, wherein the size of a convolution kernel is 3 multiplied by 3, the activation function is relu, and the filling mode is nopacking, so as to obtain a seventh output result.
And performing 1 × 1 convolution on the seventh output result, wherein the activation function is relu, and the segmentation result of the brain image data is obtained.
In this embodiment, the entire structure of the model is adjusted to seven layers. Because data acquisition of medical images is relatively difficult, privacy of the patient is involved, and the amount of data is not large. Therefore, the designed model is not suitable to be too large, the more the number of architecture layers is, the more the parameters are, the more the structure is complex, the more time is spent in calculation, and overfitting is easily caused. Maximum pooling is replaced by hole convolution. The maximum pooling operation can reduce the number of features by half, and the hole convolution is used for replacing the maximum pooling layer, so that the field of view can be increased, the loss of the features can be prevented, and more useful information can be provided for upsampling. By adding a plurality of cross-scale splicing operations and adding convolution operations in the cross-scale splicing process, the features can be better extracted without increasing the number of network layers and reducing the features.
Step (e): inputting the divided data sets into the model in the step (d) for training.
As shown in FIG. 2, FIG. 2 is a schematic diagram of the improvement of the U-Net neural network model provided by the present invention. Different from the classical U-Net neural network model, firstly the whole network structure is adjusted into seven layers, secondly the traditional maximum pooling operation is replaced by the cavity convolution, and finally all the characteristic maps obtained by each layer in the whole down-sampling process are copied, convolved and cut and respectively spliced in the up-sampling process to obtain the final result.
Illustratively, in the step (e), firstly, the result obtained in the step (c) is input into the model established in the step (d) for training, then three loss functions of binarization cross entropy, dice loss and focus loss are respectively selected to observe the influence on the training result, and the optimal training model is reserved after the training is finished.
Step (f): inputting the divided test data set into the model trained in the step (e) for testing to obtain a segmentation result.
In step (f), for example, firstly, the divided test data is input into the model trained in step (e) for testing, then the evaluation index is determined, and finally the segmentation result is obtained.
As shown in figure 1, the invention starts from the acquisition of a data set required by an experiment, then the acquired data set is preprocessed, the data set is clustered into 4 types of areas of grey brain matter, white brain matter, cerebrospinal fluid and bleeding by using an FCM method, then the skull is removed by using a morphological image method expansion and corrosion method, then an improved U-Net neural network model is established by combining cavity convolution, the divided training set is input into the model for training, three loss functions of binary cross entropy, dice loss and focus loss are respectively used for observing the influence on a training result, the optimal training result is stored, finally the divided testing set is input into the trained model for testing, an evaluation index is determined, and a division result is acquired.
In this embodiment, firstly, a fuzzy C-means clustering method is mainly used to preprocess data, and the data are clustered into 4 classes including a grey brain matter, a white brain matter, a cerebrospinal fluid and a bleeding area; then removing skull by morphological expansion and corrosion method; finally, an improved U-Net neural network model is provided by combining the cavity convolution and is used for the hemorrhagic stroke regional segmentation of the brain image, thereby facilitating the judgment and operation of the doctor operation. The accurate image segmentation improvement method can effectively reduce the difficulty of the robot operation, save the operation time and reduce the pain of patients.
It should be noted that, the steps in the method for segmenting cerebral hemorrhage based on multi-model combination according to the present invention can be implemented by using corresponding modules, devices, units, etc. in the system for segmenting cerebral hemorrhage based on multi-model combination, and those skilled in the art can refer to the technical scheme of the system to implement the steps of the method, that is, the embodiments in the system can be understood as preferred examples of the implementation method, and are not described herein again.
Those skilled in the art will appreciate that, in addition to implementing the system and its various devices provided by the present invention in purely computer readable program code means, the method steps can be fully programmed to implement the same functions by implementing the system and its various devices in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system and various devices thereof provided by the present invention can be regarded as a hardware component, and the devices included in the system and various devices thereof for realizing various functions can also be regarded as structures in the hardware component; means for performing the functions may also be regarded as structures within both software modules and hardware components for performing the methods.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (8)
1. A cerebral hemorrhage segmentation method based on multi-model combination is characterized by comprising the following steps:
step 1: acquiring brain image data to be detected;
step 2: removing a skull part in the brain image data to be detected by adopting a morphological method to obtain preprocessed data;
and step 3: inputting the preprocessed data into a target detection model, and outputting a segmentation result of the brain image data, wherein the segmentation result comprises a brain bleeding area.
2. The method for segmenting cerebral hemorrhage based on combination of multiple models according to claim 1, characterized in that before executing step 3, the method further comprises:
constructing an image dataset comprising a cerebral hemorrhage region;
dividing the image data set into a training set and a test set;
classifying the images in the training set and the test set respectively based on a fuzzy C-means clustering method to obtain intermediate image data marked with image region classification results;
removing the skull part in the intermediate image data by adopting a morphological method to obtain target data;
training the constructed initial detection model through the target data corresponding to the training set to obtain a trained model; the initial detection model is an improved U-Net neural network model established by combining hole convolution;
and testing the trained model through the target data corresponding to the test set, and if the test is passed, obtaining the target detection model.
3. The method for segmenting cerebral hemorrhage based on multi-model combination according to claim 2, wherein the constructing the image dataset comprising the cerebral hemorrhage region comprises:
and screening the images containing the cerebral hemorrhage area, screening the images with the image size, the image definition and the image integrity meeting the requirements, and adding the images into the image data set.
4. The method for segmenting cerebral hemorrhage based on combination of multiple models according to claim 2, wherein the images in the training set and the test set are classified respectively based on a fuzzy C-means clustering method to obtain intermediate image data labeled with image region classification results, and the method comprises the following steps:
based on a fuzzy C-means clustering method, clustering is respectively carried out on the gray value of each pixel of the images in the training set and the testing set, and intermediate image data marked with four types of region classification results, namely a grey brain matter region, a white brain matter region, a cerebrospinal fluid region and a bleeding region, are obtained.
5. The method for segmenting cerebral hemorrhage based on combination of multiple models according to claim 2, wherein the skull portion in the intermediate image data is removed by a morphological method to obtain target data, and the method comprises the following steps:
and adjusting the expansion parameter and the erosion parameter of the intermediate image data until the skull part in each image is removed.
6. The method for segmenting cerebral hemorrhage based on combination of multiple models according to any one of claims 2-5, characterized in that the combination of hole convolution to establish the improved U-Net neural network model comprises a seven-layer convolution structure, wherein:
the first layer is used for: performing two convolutions and a cavity convolution on input data, wherein the size of a convolution kernel is 3 multiplied by 3, an activation function is relu, stride is 2, and a filling mode is nonpadding to obtain a first output result;
the second layer is used for: performing two convolutions and a cavity convolution on the first output result, wherein the convolution kernel size is 3 multiplied by 3, the activation function is relu, stride is 2, and the filling mode is nonpadding, so as to obtain a second output result;
the third layer is used for: performing two convolutions and a cavity convolution on the second output result, wherein the convolution kernel size is 3 multiplied by 3, the activation function is relu, stride is 2, and the filling mode is nonpadding, so as to obtain a third output result;
the fourth layer is used for: performing two convolutions on the third output result, wherein the size of a convolution kernel is 3 multiplied by 3, an activation function is relu, and a filling mode is nonpadding to obtain a fourth output result;
the fifth layer is used for: performing convolution and clipping operations on the copied first output result, the copied second output result and the copied third output result, then performing splicing and upsampling operations, wherein the size of a convolution kernel is 3 multiplied by 3, and the size of upsampling is 2 multiplied by 2, so as to obtain a fifth output result;
the sixth layer is for: performing two convolution and one upsampling on the copied fourth output result and the fifth output result, wherein the size of a convolution kernel is 3 multiplied by 3, the size of an activation function is relu, the size of upsampling is 2 multiplied by 2, and the filling mode is nopadding to obtain a sixth output result;
the seventh layer is for: performing two convolutions on the copied fifth output result and the copied sixth output result, wherein the size of a convolution kernel is 3 multiplied by 3, an activation function is relu, and a filling mode is nonpadding to obtain a seventh output result;
and performing 1 × 1 convolution on the seventh output result, wherein the activation function is relu, and the segmentation result of the brain image data is obtained.
7. The method for segmenting the cerebral hemorrhage based on the combination of multiple models according to any one of claims 2-5, characterized in that when the constructed initial detection model is trained through the target data corresponding to the training set, three loss functions of binary cross entropy, dice loss and focus loss are adopted to observe the influence on the training result.
8. A multi-model combination-based cerebral hemorrhage segmentation system, which comprises a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program stored in the memory, the multi-model combination-based cerebral hemorrhage segmentation method according to any one of claims 1 to 7 is executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011235586.1A CN112348796B (en) | 2020-11-06 | 2020-11-06 | Cerebral hemorrhage segmentation method and system based on multi-model combination |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011235586.1A CN112348796B (en) | 2020-11-06 | 2020-11-06 | Cerebral hemorrhage segmentation method and system based on multi-model combination |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112348796A true CN112348796A (en) | 2021-02-09 |
| CN112348796B CN112348796B (en) | 2024-01-30 |
Family
ID=74429942
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011235586.1A Active CN112348796B (en) | 2020-11-06 | 2020-11-06 | Cerebral hemorrhage segmentation method and system based on multi-model combination |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112348796B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116245951A (en) * | 2023-05-12 | 2023-06-09 | 南昌大学第二附属医院 | Brain tissue hemorrhage localization and classification and hemorrhage quantification method, device, medium and program |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109035252A (en) * | 2018-06-29 | 2018-12-18 | 山东财经大学 | A kind of super-pixel method towards medical image segmentation |
| WO2020007277A1 (en) * | 2018-07-05 | 2020-01-09 | 北京推想科技有限公司 | Cerebral hemorrhage amount calculation method based on deep learning |
| CN111724397A (en) * | 2020-06-18 | 2020-09-29 | 上海应用技术大学 | Automatic segmentation method for bleeding area of craniocerebral CT image |
-
2020
- 2020-11-06 CN CN202011235586.1A patent/CN112348796B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109035252A (en) * | 2018-06-29 | 2018-12-18 | 山东财经大学 | A kind of super-pixel method towards medical image segmentation |
| WO2020007277A1 (en) * | 2018-07-05 | 2020-01-09 | 北京推想科技有限公司 | Cerebral hemorrhage amount calculation method based on deep learning |
| CN111724397A (en) * | 2020-06-18 | 2020-09-29 | 上海应用技术大学 | Automatic segmentation method for bleeding area of craniocerebral CT image |
Non-Patent Citations (2)
| Title |
|---|
| 王海波;李雪耀;: "基于FCM聚类算法的颅内出血CT图像分割", CT理论与应用研究, no. 02 * |
| 王海鸥;刘慧;郭强;邓凯;张彩明;: "面向医学图像分割的超像素U-Net网络设计", 计算机辅助设计与图形学学报, no. 06 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116245951A (en) * | 2023-05-12 | 2023-06-09 | 南昌大学第二附属医院 | Brain tissue hemorrhage localization and classification and hemorrhage quantification method, device, medium and program |
| CN116245951B (en) * | 2023-05-12 | 2023-08-29 | 南昌大学第二附属医院 | Brain tissue hemorrhage localization and classification and hemorrhage quantification method, device, medium and program |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112348796B (en) | 2024-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mateen et al. | Exudate detection for diabetic retinopathy using pretrained convolutional neural networks | |
| Gecer et al. | Detection and classification of cancer in whole slide breast histopathology images using deep convolutional networks | |
| CN110232383B (en) | Focus image recognition method and focus image recognition system based on deep learning model | |
| CN109035255B (en) | Method for segmenting aorta with interlayer in CT image based on convolutional neural network | |
| WO2019200753A1 (en) | Lesion detection method, device, computer apparatus and storage medium | |
| Veena et al. | A novel optic disc and optic cup segmentation technique to diagnose glaucoma using deep learning convolutional neural network over retinal fundus images | |
| CN108198184B (en) | Method and system for vessel segmentation in contrast images | |
| Tian et al. | Multi-path convolutional neural network in fundus segmentation of blood vessels | |
| CN111951221A (en) | Glomerular cell image identification method based on deep neural network | |
| CN111724397B (en) | Automatic segmentation method for craniocerebral CT image bleeding area | |
| CN112862805B (en) | Automatic auditory neuroma image segmentation method and system | |
| Castro et al. | Convolutional neural networks for detection intracranial hemorrhage in CT images. | |
| Kumar et al. | Image processing in diabetic related causes | |
| CN116188485A (en) | Image processing method, device, computer equipment and storage medium | |
| CN102419864B (en) | Method and device for extracting skeletons of brain CT (computerized tomography) image | |
| CN112348796B (en) | Cerebral hemorrhage segmentation method and system based on multi-model combination | |
| Kyaw | Computer-Aided Detection system for Hemorrhage contained region | |
| Anitha et al. | A segmentation technique to detect the Alzheimer's disease using image processing | |
| CN112102351A (en) | Medical image analysis method and device, electronic equipment and readable storage medium | |
| Azli et al. | Ultrasound image segmentation using a combination of edge enhancement and kirsch’s template method for detecting follicles in ovaries | |
| CN112862785B (en) | CTA image data identification method, device and storage medium | |
| CN111951219B (en) | Thyroid eye disease screening method, system and equipment based on orbit CT image | |
| CN114926396A (en) | Mental disorder magnetic resonance image preliminary screening model construction method | |
| CN111462067B (en) | Image segmentation method and device | |
| Sameer et al. | Brain tumor segmentation and classification approach for MR images based on convolutional neural networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |